The Agent Skills Directory

[COMMAND_EXECUTION] (LOW): The skill includes a bash script find-polluter.sh and various CLI examples (e.g., npm test, security list-keychains). These are standard diagnostic tools for software development and debugging. The script executes npm test on files discovered via a user-defined pattern, which is intended behavior for a developer-centric utility.
[DATA_EXFILTRATION] (SAFE): No network-based exfiltration patterns or hardcoded credentials were found. Examples provided for environment variable checking (e.g., ${IDENTITY:+SET}) are designed to verify existence without exposing actual secret values.
[PROMPT_INJECTION] (SAFE): The skill contains internal behavioral constraints ("The Iron Law") to prevent the agent from rushing to fixes, but these do not attempt to override system-level safety protocols or jailbreak the model.
[REMOTE_CODE_EXECUTION] (SAFE): There are no patterns involving the download and execution of external scripts (e.g., curl | bash). All code components are local and transparent.

systematic-debugging